Generative AI has become the business world’s obsession. Leaders are pouring resources into pilots, hoping for breakthrough returns. Yet the numbers are sobering. According to MIT’s GenAI Divide: State of AI in Business 2025 report, 95% of AI pilots are failing to deliver a measurable impact. Only about 5% accelerate revenue in a meaningful way.
Such a deeply negative trend might be expected to kill off future projects, but the risk of falling behind on AI is outweighing cold economics. Many business leaders believe they simply have to get this right.
What is the factor that makes such a huge difference between success and failure? MIT found that success is less about the technology and more about how organizations approach implementation. As their researchers put it, the gap is “not the quality of the AI models, but the ‘learning gap’ for both tools and organizations.” In other words: the fundamentals aren’t in place.
The MIT study is based on 150 interviews with leaders, a survey of 350 employees, and an analysis of 300 public AI deployments. Its central finding is clear: most pilots stall because businesses are seduced by flashy use cases yet they neglect investment in the infrastructure and culture to support them.
Aditya Challapally, the report’s lead author, summed it up: “Some large companies’ pilots and younger startups are really excelling with generative AI … because they pick one pain point, execute well, and partner smartly.” By contrast, most enterprises misallocate resources, over-indexing on front-office tools while ignoring back-office automation – yet that is where the biggest ROI actually lies.
The data shows that more than half of generative AI budgets go to sales and marketing tools, but the highest returns come from areas like document automation, process streamlining, and eliminating outsourcing costs. This is exactly where foundational investments in data and platforms pay dividends.
The Fundamatic white paper highlights the same story, through the lens of asset allocators. Investment institutions, from foundations to pension funds to family offices, struggle under a relentless deluge of documents and data. Analysts often spend 75–80% of their day on manual processes: logging into portals, downloading reports, renaming files, and extracting relevant data.
This isn’t just inefficient; it’s corrosive. Natural human error is unavoidable. Mistakes creep in, decision-makers can’t fully trust the data, and talented employees burn out. Growth becomes constrained because operations teams can’t keep up.
The fix is not another experimental pilot. It’s putting in place a robust foundation for automating document and data input, ensuring that it takes place quickly, reliably, and generally unnoticed. Fundamatic calls this philosophy Quietware™: software that reduces distractions, handles repetitive work, and ensures a flow of trusted data without constant human oversight.
Taken together, MIT’s research and Fundamatic’s alts industry experience point to the same conclusion: success in AI comes from fixing the fundamentals first.
Here’s what that looks like:
The MIT report makes a blunt observation: “The 95% failure rate for AI solutions represents the clearest manifestation of the GenAI Divide.” That divide isn’t inevitable. It’s the outcome of poor methodology – ignorance about the data foundation GenAI needs to be effective, focusing on outcomes without fully considering inputs, failing to appreciate how AI tools will integrate with teams and other technologies, and misguided incentives that allow AI to be seen as a threat rather than a workplace enhancer.
Companies that start with strong data foundations, smart partnerships, and workflow-level integration are the ones that cross the divide.
Fundamatic’s experience with alts firms reinforces this. Operations organizations that have implemented Vault, its core product for automating fund document ingestion and classification, not only reduce manual workloads but also unlock analyst capacity for strategic innovation. Ops leaders find that their team is seen as adding value, no longer just a necessary cost center.
After some time, it becomes clear to the wider business – including the investment side of the house – that data quality has improved. Fund data that is clean, complete, and up-to-date becomes ‘trusted data’ – and that is the foundation for compounding ROI. Some Fundamatic clients build on this with Analyze, a product that enables investment teams to explore and extract insights that improve portfolio management and performance.
Executives eager to capitalize on AI must resist the temptation to jump straight into use cases. Pilots without platforms are like skyscrapers built on sand: they look impressive but won’t withstand the demands of reality. Selecting the right partner is also crucial. Of course AI vendors promise automation, but not all are able to deliver it with lean, efficient technology. Beware of outsourcers masquerading as pure technology – a time-limited test can weed out those they rely on hidden people.
The lesson from both MIT and Fundamatic is clear: get the fundamentals right first. That means quiet automation, trusted data, and platforms that integrate deeply with your business processes. Only then will AI deliver on its promise. And only then will you join the rarified minority of organizations that see exponential returns.
Download the full white paper to explore how Quietware™ and trusted data can reshape your operations from the ground up.